Short-Term Traffic Flow Prediction Based on Multilinear Analysis and k-Nearest Neighbor Regression

Prevailing short-Term traffic flow prediction models concentrate on using black-box type of artificial intelligence (AI) algorithms without explicit knowledge of traffic flow data. In this paper, a novel short-Term traffic flow prediction method Ml-k-NN was developed using multilinear analysis. The model recognizes the lane flow distribution within the traffic flow data and uses the k-nearest neighbor to predict traffic flow. The proposed multilinear analysis technique employs a dynamic tensor form of traffic flow data and uses tensor decomposition to combine several characteristics of traffic flow data. With the tensor decomposition, we can not only find the spatial-Temporal information and lane distribution of traffic flow pattern, but also acquire the short-Term traffic prediction by applying the k-nearest neighbor method on the generated features. Experiments on real traffic data acquired from 10 locations on 4-lane freeway are provided to validate and evaluate the proposed approach. Experimental results show that the proposed method has the promising performance in predicting traffic flow.